… The UK government-backed AI Security Institute reports that during a series of cybersecurity evaluations, Anthropic’s Claude Mythos 5 and OpenAI’s GPT-5.6 Sol both took “autonomous, unsanctioned action on the live internet,” including an instance where an agent attempted to upload malicious code to… …
… The malicious package was downloaded and installed by 15 real-world companies, including a security firm, Anthropic admitted. …
… According to Anthropic, these checks may be used to manage user accounts, prevent fraud, investigate misuse and security incidents, and ensure compliance with their terms of service. …
…The "good" news about this Claude incident is that it involved only shared Claude chats. That is, Claude conversations that you've specifically made available to others via a public link. The…